Chen, G., Dong, S., Shu, Y., Zhang, G., Sesay, J., Karlsson, B., Fu, J., & Shi, Y. (2024). AutoAgents: A Framework for Automatic Agent Generation. arXiv.

The paper titled “AutoAgents: A Framework for Automatic Agent Generation” by Guangyao Chen, Siwei Dong, Yu Shu, Ge Zhang, Jaward Sesay, Börje Karlsson, Jie Fu, and Yemin Shi (2024) introduces a framework called AutoAgents designed to dynamically generate and coordinate specialized agents for task-solving. Unlike existing approaches that rely on predefined agents, AutoAgents adapts to different scenarios by generating agents based on the content and requirements of the task at hand. The framework involves two key stages: a Drafting Stage, where a customized agent team and execution plan are created through collaborative discussions among predefined agents, and an Execution Stage, where these agents refine and execute the plan collaboratively. Experiments demonstrate that AutoAgents produces more coherent and accurate solutions compared to existing multi-agent methods, highlighting the importance of dynamic agent generation and collaborative refinement in handling complex tasks. The work suggests that AutoAgents can significantly enhance the effectiveness of AI in complex problem-solving by facilitating the synthesis of diverse knowledge and expertise.

https://arxiv.org/abs/2309.17288